Talk:Reduced row echelon form: Difference between revisions
→New test cases
Line 173:
: It's been ages since I have thought about this task. If we get differing results, how do we determine which result is "better". --[[User:Rdm|Rdm]] 16:35, 16 January 2012 (UTC)
:: I think matrix three is designed to show up floating point errors. Using the optional left argument for the <code>gauss_jordan</code> verb used in the J solution hints at that and using extended precision gives the desired answer. --[[User:Tikkanz|Tikkanz]] 15:31, 26 January 2012 (UTC)
<lang j> 1e_14 gauss_jordan mat_3
1 2 0 0 3 4
0 0 1 0 0 _1
0 0 0 1 0 1.4803e_16
0 0 0 0 0 0
0 0 0 0 0 0
gauss_jordan x: mat_3
1 2 0 0 3 4
0 0 1 0 0 _1
0 0 0 1 0 0
0 0 0 0 0 0
0 0 0 0 0 0
</lang>
| |||